Better Offline How To Argue With An AI Booster, Part Two
70 snips
Sep 11, 2025 In this discussion, listeners dive into the complex economics of AI investments, exposing substantial losses that contrast sharply with traditional tech companies. The conversation debunks myths about the cost and viability of AI, highlighting emerging challenges in GPU affordability and innovation. A critical perspective is offered on the hype surrounding AI technologies, urging a more realistic understanding of their societal impact. The financial landscape is dissected, shedding light on misconceptions regarding open-source versus proprietary AI solutions.
AI Snips
Chapters
Transcript
Episode notes
GPUs Aren't The New Fiber
- GPUs are a centralized, depreciating capital asset rather than a broadly useful infrastructure like fiber optics.
- That means huge pools of GPUs may sit idle if no new mass-market use case appears.
Secondhand GPUs Are Already Cheap
- Ed recounts finding A100 GPUs for $2,000–$3,000 on eBay and H100s for about $30,000 used versus $45,000 retail.
- He uses these examples to show GPUs are already widely available and depreciating.
AI 2027 Is Speculative Fanfiction
- AI 2027 is framed as speculative fanfiction, not proof of an imminent self-improving agent.
- The report uses scary language and credentials to persuade credulous readers, not to demonstrate feasibility.